 
|
Audio Asylum Thread Printer Get a view of an entire thread on one page |
For Sale Ads |
|
|
Audio Asylum Thread Printer Get a view of an entire thread on one page |
For Sale Ads |
144.131.79.142
In a very comprehensive and detailed article, published in Japan in 1999 and reproduced in Glass Audio the following year, Satoru Kobayahi described a 300B SE stereo amp that uses Plitron toroidal OP and power transformers.I was interested to see that he used a power MOSFET series regulator in place of the smoothing choke in what would have been a conventional CLC pi-filter.
Essentially, his PS consisted of: toroidal power transformer - SS bridge - 100uF smoothing cap - MOSFET series regulator - 100uF reservoir cap.
For regulation, he had well-bypassed stacked zeners (3 x 150 v) from the MOSFET gate to ground. According to the test voltages written on his schematic, he dropped 33 v across the MOSFET.
His initial objective, in replacing a choke with a MOSFET, was to save money and reduce weight. However, he also reported that the ripple was much less than he could have achieved with a 10H choke. In view of Satoru-san's reputation as a designer and builder, I'm inclined to think that if it's good enough for him then it's good enough for me!
Follow Ups:
I've recently been doing a bunch of experiments along these lines. The upshot is that it IS possible to get very good sounding results using high voltage regulators, but not just any old regulator will do.I tried a bunch of different designs I could find on the web and non of them were doing all that great until I got to the last group. I tried series regulators and shunt regulators, tube, BJT and MOSFET. My previous best result was a shunt regulator driving a series regulator (I got VERY good line rejection from the shunt, but the output impedance was not as low as I liked, the series had very low output impedance, but not so great line rejection). This combo sounded pretty good but was very complicated and needed large heatsinks, it wound up wasting a lot of power.
What changed this was a bunch of threads on regulator design over at diyhifi.org, this got me really analyzing the designs and finding the good and pad parts. I wound up with a modification of the infamous Gary Pimm CCS to turn it into a very nice series high voltage regulator. The sonic results of this are way better than anything else I've ever tried. The line rejection is quite high and very low output impedance from very low frequencies to well above the audio range.
So far I have tried this on B+ and screen supplies for several lower power applications, (12 watts max) primarily because I don't have a big enough heatsink yet to try it on some higher power amps. Unfortunately I'm a little afraid to try it on the big 813 amp, the 700V B+ is a little too close for comfort to the max rating of 1000V MOSFETs. On preamps its absolutely stunning.
Now on the aspect of choke input, cap input etc it gets rather interesting. It IS possible to use a cap input filter with a small cap, say 2-4uf that still keeps the diodes and transformers etc cunducting for a good portion of cycle. I did some experiments with large C, small C and critical inductance choke input and mapping out the junk they spewed into the environment. As has been mentioned the large C sprays a lot of stuff around inside the chasis and into the power cord etc. The choke input gets rid of almost all of that. The small C was almost as good as the choke input. At one point I was trying .47uf, I could not tell any difference with this.
The problem with the small C is a very large ripple after the C. So the transformer voltage has to be high enough that the bottom of the ripple is sufficiently high above the droput voltage of the regulator that you always stay above regulation no matter what the load is. This also means that you are converting that entire portion of ripple above the output voltage to heat, so you need a good size heatsink. So yes you can get away without a big choke, but then you need a big heatsink. Its all about tradeoffs.
Of course you can use a choke input filter WITH the regulator, there is nothing to prevent that. Just go with a simple LC first stage followed by the regulator. The single stage LC still has pretty high ripple but the regulator takes care of that.
I haven't finished all the experiments yet, but things are leaning towards the conclusion that results from using this regulator are sonically better than what I got with the low DCR, low H C etc design. As far as I could tell what that supply gave me was a quite low output impedance over the whole audio range with very fast recovery for large transients. The regulator design does the same thing, very low output impedance over the whole audio range, and no transient issues at all as long as the supply driving it has the power to handle the largest transients.
So it looks like a choke input design that I would normally not like because of fairly high output impedance in some parts of the spectrum can be used very effectively with a GOOD regulator. Of course you still need a supply that can handle the largest current spikes without dropping below the regulator dropout voltage. This still needs pretty good size chokes and transformers etc. Or you could go with a low C input and live with the somewhat larger heatink.
I'm currently in the process of trying both of these out on my preamp. It will be awhile though before I get both built and tried out. I really want to try this out on my big amp but I have not yet come up with a good way to handle the voltage to my satisfaction.
Oh yeah the schematic! Right now its just a bunch of scrawls on a pad. I'm going to try and get this on the computer soon and post it here. My computer has been down for a week so I haven't gotten it done yet. I wasn't quite ready to post about all this, but this thread was just too tempting!
John S.
Hi John, that's very interesting and I'm looking forward to seeing your designs."Unfortunately I'm a little afraid to try it on the big 813 amp, the 700V B+ is a little too close for comfort to the max rating of 1000V MOSFETs."
In a series MOSFET regulator, this shouldn't be a problem, should it? (unless you short the reg output accidentally to ground, that is!) I suppose if you were ue=sing a MOSFET as an error amp in a series rgulator, then it would be a concern although, even then, you could safeguard yourself by using a high voltage zener or VR tube in the source-to-ground circuit of the error amp. At least with MOSFETs you don't have heater-cathode voltage limits to worry about!
so I kinda feel guilty adding to it but I'd like to share a couple of experiences along the lines of what you describe. Years ago I modded some ARC Ref 600 PP amps. One of the first things I did was to add a choke to a simple SS bridge / big C, HT PS. The gain in smoothness was not insignificant. I went on to add a Mosfet regulator to stabilise the HT voltage (for other reasons) and to be honest, it didn't make a great sonic difference.Down the track I built some 845s and thought I'd try a simple Mosfet Regulator again. I built one exactly as you've described and for the heck of it I wanted to see the difference the regulator made with and without a choke. The choke always helped, though by a smaller margin with the regulator in circuit.
One big warning ... actually a couple if you try this. First, make sure that you ramp the voltage up slowly or you'll wonder why those expensive Mosfets keep blowing up … no matter how much current they'll handle. How do I know this? Well I almost went deaf and blind in the process! BTW, bipolars aren’t any better. Second, you have to take the Mosfet’s input capacitance into account and drive it with a low impedance or you'll lose the highs. Alternatively, (but not ideal) you can bypass the Mosfet with a small C. You'll get back the highs but you'll forego the benefit of HF regulation.
Rgds,
Naz
. . . I don't really see it like that. It seems more to me like a bunch of concerned and experienced people sharing their knowledge, in the hope of preventing someone from being misled into making a regrettable mistake. Actually, I wasn't really supporting the idea, just interested in it, since it came from a 'usually reliable source'. It sure led to a lot of responses, though, so that's good!I take it that your suggestion to ramp up the voltage slowly would not be satisfied by an RC time constant on the gate, to allow it to reach the reference voltage from the zeners slowly? That may explain why MOFET serial regulators usually have a series resistor of ~100 ohms, and why Satoru-san uses a delay relay to bypas a resistor in series between the bridge and the smoothing cap.
> > I take it that your suggestion to ramp up the voltage slowly would not be satisfied by an RC time constant on the gate, to allow it to reach the reference voltage from the zeners slowly? < <This can be done but it means using a large C in order to achieve a high enough time constant with a low enough value of R. The time constant needs only to be large enough to limit inrush current caused by having to charge the OP Caps.
> > That may explain why MOFET serial regulators usually have a series resistor of ~100 ohms> >
The series (100R) that people hang off the gate is to prevent HF oscillation ... similar to a grid stopper. It's also important to place a zener from S to D to limit potential over-voltage to a safe level.
BTW, there was a comment made by JS about the voltage rating of the Mosfet. If the INPUT voltage is ramped up slower than the regulator's ability to charge the output C then barring a short or sudden instant VERY heavy load on the PS the voltage rating needn't be all that high because the differential across any of its terminals will never be greater than the input V minus the OP V at any given moment. Staging the input voltage is also a reasonable technique here.
Rgds,
Naz
"The series (100R) that people hang off the gate is to prevent HF oscillation"
I know, but I didn't mean in series with the gate, I meant in series with the source. I've sometimes seen it described as a 'safety resistor'. I suppose it's to limit the current if the output is short-circuited (as it would be, for a short time after switching on, when the following reservoir cap has to be charged up).The time constant I had in mind was a 2.2Meg resistor, from the top of the zener (or VR tube, if you prefer) stack to the gate stopper, with a 2uF cap from the gate stopper to ground. Since the MOSFET has such a high gate impedance, a very high value of resistance can be used if required. This would cause the gate voltage to rise slowly, say about 10 seconds from 0v to the reference voltage.
Sorry about the misinterpretation but I've never seen a 100R used in series with the source. As to its purpose I'm sure it's as you suggest but unless it's being used in a low current application I think it would be detrimental to sound.
> > The time constant I had in mind was a 2.2Meg resistor, from the top of the zener (or VR tube, if you prefer) stack to the gate stopper, with a 2uF cap from the gate stopper to ground. Since the MOSFET has such a high gate impedance, a very high value of resistance can be used if required. This would cause the gate voltage to rise slowly, say about 10 seconds from 0v to the reference voltage. < <The time constant is OK and even lower will do but IME the high value R can affect the HF response if a good quality cap is not used.
Naz
A resistor after the source would negate the regulation. Put it before the drain, though, and it has no effect on reg'n provided that it's not too big.
I first tried this when I added a series reg on my PP ref 600s. I kept blowing up huge Mosfets and even Bipolars, so I thought I'd try a resistor in series with the drain. It absolutely annihilated Dynamics and the amp sounded dead ... unlistenable (I know that’s not a word)! Removed the R and the whole thing came back to life! It obviously just slows things down too much.Naz
Hi
I read an old electronics book where resonance chokes where sugested as a soulution for this dilemma.
Not just the regular resonance choke design, but to use a regular PSU transformer in 1:20 ratio (aprox. 110V:6V in US or 230V:12V in EU)
The secondary (6V or 12V) is used in the amplifier filter for really low DC resistance (aprox. 1-2Ω).
The primary (110V or 230V) is connected as a resonance circuit with a cap in paralell with the primary winding (The resonance freq. in found where L and C reaktance is equal. (L- and C-resistance neglected).
The book sugested a futher adjustable cap in parallel for fine adjustment of the resonance peak to 2 x mains freq = 120Hz(US) or 100Hz(EU) .
This could also be done in another way (-cheaper), by connecting a potentiometer i series with the fixed cap.
This is because the resonance freq. is actually where the Impedance of the L and C, so by adding a variable resistance to the C-leg of the resonance circuit, the Q-value vill vary somewhat, but most, the resonance freq. vill be adjustable.Remember:
The DC current trough the secondary is only going through small amount of wire-turns, which make a regular x-former ok without an airgapped core.
The impedance will be reduced from primary to secondary side, but as the circuit is a resonance peak, the impedance is really high at 120Hz(US) or 100(EU).
TheThe circuit has only a high impedance at a very narrow band (easy to calc. by using the Q-value), so perhapps a second fiterstage (smaller hi mu choke) should be used to eliminate hi freq. ripple.
I hope I made myself understud..
I just would like to hear your comments about it as an alternative to the one sugested in the tread.Regards
Pix ;-)
Sweden
Hi Ray,You might be interested in reading about the Minimal Reactance Power supply.
Dave
Dave,Problem is I use Class AB1, so the idea of shunt regulation doesn't appeal to me, even though I realize it's quieter.
...will be lost if there is no choke.And this point is that a PROPER choke input filter doesn't take pulses from the AC line at the peak of the waveform, but takes an almost a pure AC signal - hence the AC wiring, transformer, diodes and assocciated circuitry don't splash intense pulses of RF around the inside of the amp at 100 or 120Hz and all harmonics up to daylight!
A following regulator will/should kill the B+ ripple for certain, but so what if the sensitive input circuitry is being bathed in wideband pulsed RF?
A second point is that these pulses are present in the powertransformer's magnetic field - and therefore appear on the heater lines, and on any other voltage made from the same transformer that is driving a cap input filter!
The same thing happens with a cap input DC heater supply for a preamp/phono stage - this can make HUGE pulses which lay themselves over the B+, even if that's using a choke input filter.
Personal experience...not just theory!
Regards, Allen
Hi Allen,I understand your point. Since the main purpose of Satoru Kobayashi's PS design was to avoid the use of a heavy and expensive choke, I wonder if reducing the size of the input C before the MOSFET could help to minimize the spikiness to any useful extent? Or would a small inductor < < 1Hy before the input C, as proposed by Jeff and experimented with by Henry, help to neutralize the RF? BTW, did you find this to be a problem in your own PP-1C?
Huh? the schema I have of his amp (GA #4/1999)uses a 300mA/10H choke - which isn't small/light by anyone's standards, followed by a MOSFETregulator for the B+.Are we talking about the same thing?
My PP-1Ccircuit was always intended to be a very low cost unit to be sold in malaysia as a kit. In it's original form it even used a 25 watt toroidal power transformer as an OPT.
Later versions used good James OPTs and became something rather nice - but when it grew into the PP-2C with a choke input filter - it stepped up several levels of perfomance with no other circuit or parts changes.
QED!
Regards, Allen
Allen, the design I'm talking about can be seen at the link below. There is no choke and in the text Satoru-san explains that he wanted to avoid using one, to save weight and cost. And thanks for explaining your findings with the PP-1C.
...both power and output traffos, even a high end Lundahl choke would add a very small percentage to both the cost and the weight - and minimally to the already gigantic (to me at least) size.I like my chiropractor, but hate the time between a strain and an open date at his office - hence I only ever build monoblocks!
Never saw that article - only the huge P-P one I referenced.
Regards, Allen
"even a high end Lundahl choke would add a very small percentage to both the cost and the weight - and minimally to the already gigantic (to me at least) size."
Well, I tend to agree, but I'm just quoting what he said in his article. Maybe, with all the cost, space and weight he'd already incurred, he felt that a smoothing choke would have been the last straw?As I explained to Tre', I wasn't crusading against the use of chokes, just interested in this alternative approach. From your replies and others, it seems a choke-input filter (a real one!) is a hard act to follow. Thanks for your interest.
That is my experience, without any reservations, across 30 years of actual work.Talked at length with Tim de Paravancini at last year's European Triode Festival, and we found we both agreed on the importance of choke input filters. But he's even more emphatic about this than me. I told him that in the "TubePreamp CookBook" I write that the choke should be at least the same physical size/weight of the power transformer - and he contradicted me: "No - it should be twice the size/weight!"
Regards, Allen
Hi Allen,Our mutual friend, the late Bob Fulton, used to tutor me in about 1973 on the use of " twenty ohms or less DCR for the chokes " which IS a large unit. Now - in 2007, I have changed to ten ohms or less DCR, and under 1/2 a HY, and Low C.
Even if this is may not intuitive to you, it is worth a build and listen Allen!! Use a couple Triad C-40Xs and two 50 uF ( or 40 uF ) caps, high quality, and heavy AWG high-quality hook-up wire. L/C/L/C.
Have fun listening to that! It is "something else" on dynamically recorded music, especially when all the musicians play at once !!
Refreshingly good to hear. A whole new performance level, one that we never even had before in audio.
The " proof " many of you seek of this build approach is very simple, ........ it is in the listening Allen.
Cheers,
Jeff Medwin
Ray,IMO, the place for a high current/low DCR RF choke is AFTER the 1st filter cap. and in front of the regulator. The "hash" is associated with the triangular wave form cap. I/P filters ALWAYS exhibit.
Eli D.
How right you are. I tried to show this with screen shots, from the PSUD, of the huge charging spikes present in a supply with an input choke that does not meet critical inductance.No one seemed to notice. I know people respect you so I am showing this screen shot again in hopes that people will take notice.
A choke input supply with too small a choke is not a choke input supply at all. It will not have the regulation of a choke input supply and it WILL have the spikes of a cap input supply.
This circuit is drawing about 80ma. but we have large (over 350ma.) charging current spikes. More important than the amplitude is the shape. Those sharp spikes will, as you say, create RF that radiates through out the amp.
With a proper choke in place and the circuit adjusted to draw the same 80ma. DC, the charging currents are held down to just over 25ma. peak and are Sinusoidal, no spikes = no RF.
Tre'
Have Fun and Enjoy the Music
"Still Working the Problem"
Okay, I'm convinced! I wasn't crusading against choke input filters, just asking a question, which now seems to have been well and truly answered. Thanks, everybody!
Solve one problem, create another !! Boo and G-d bless Ivan303 who I've plagerized.JM
In general, how many Henries do you need to really get critical inductance? I'm guessing it depends on current draw?
The formula is printed in the Radiotron Designer Handbook 4th edition, and depends on both the load presented to the power supply RL = voltage of power supply/current drawn by load, and the AC frequency. For a 50 Hz supply,L = RL/940
For a 60 Hz supply,
L = RL/1130.
These are approximate values to achieve continuous current flow according to the RDH. If I read the text correctly, this also suggests that below the critical inductance the current flow is not continuous during the whole cycle.
then too much should be just about right!I don't know what Jeff's doing or hearing, but it makes NO sense from any understanding I have - no more than his "crazy" grid stopper concepts.
And this isn't a critque of Jeff as such - I just wish his ideas came with some understanding.
Regards, Allen
Can't speak for Jeff, but he's expressed some, well, I would call it comtempt, for the RDH in some past posts. It seems his answer for any criticism is "just listen to it."
What Henry showed us, not Jeff, was that the right combo of small first choke (L1) and small first cap (C1) made it possible to charge C1's voltage to more than the normal max of 1.4 times the secondary RMS value. When a more "normal" LC stage followed it, regulation proved to be very good. Tuning is rather touchy (load dependent too). Personally I was troubled by the very high voltages developed in the first choke but Henry felt it was a limitation on Duncan's sim. I have not bothered to build it nor model anymore. Call me old school but I have no problem with a regular choke input filter.
Russ
Choke value in henries = ((voltage /current) + resistance of choke))/1000Tre'
Have Fun and Enjoy the Music
"Still Working the Problem"
At any given rail voltage, there is a minimum (critical) inductance that's required for proper choke I/P operation. Frankly, I don't know the formula for critical inductance.Once the critical inductance criterion is satisfied, a minimum current draw (the critical current) is needed for proper operation. A good approximation for the critical current, in mA., is V/L. When a bleeder resistor is used to ensure the critical current gets drawn, it turns out that 1 KOhm of resistance is appropriate for each Henry of inductance present in the 1st position.
Eli D.
Allen,That's a great way to explain the need for a choke-input. I started using choke-input supplies for B+ (most with subsequent regulators) almost exclusively in years past (back when I had time to build things). The elimination of those huge magnetic field spikes adds a certain ease to the sound. And no matter how robust one's ground wiring is, huge current spikes will create voltage pulses in the ground system.
_____________
Brian
...like a stereo 6AS7 PP amp. 300mA idle, God knows how much current at transient peaks. Getting a 10H 500mA choke, or a two pair of 10H 200mA chokes, into a sane sized chassis wouldn't be fun. The PT would already be enough of a PITA. I'm thinking a 100VA control transformer, 120V input, two 240V outputs (series for some 12AX7 SRPP drivers, parallel for the 6AS7 B+).
Weight and cost savings, most certainly. Space saving is limited by the fact that the MOSFET will be regulating the full current of the amp and will need a good heatsink.
I'm really interested in pursuing this option. Are there any sites or written works you could point me towards to get a crash course in SS HV regulation?Alternatively, would a 4H 200mA 65-ohm choke after a SS bridge have sufficient regulation? The PT would be a 105VA 120 -> 2x240V control transformer, and the load is 150mA idle per channel for a PP 6AS7. 200V B+ is what I'm shooting for. Sorry for threadjacking :(
Sorry, I don't have any more info. This is all new to me too, that's why I wanted to discuss it here. BTW, I saw a pile of posts from Jeff and others earlier on, but they've disappeared now!
Ray,I think the concept has merit, but can be improved on. A large capacitance at the I/P of a PSU filter introduces ripple freq. overtones into the B+ rail. A choke seems better to me for killing that sort of crud than a regulator is. However, if the I/P capacitance is kept reasonably small, the ripple fundamental will be dominant and the regulator WILL kill it. Set things up along the lines used with vacuum rectification. Place the bulk of the energy storage at the end of the filter.
Eli D.
Yes, Eli, I must admit that I prefer a fairly small smoothing cap (directly following the rectifier) and a much larger reservoir cap (after the filter). I'm talking about 50uF and 470uF respectively. Also, I would prefer to ramp up the voltage slowly by using a high value resistor, between the reference zeners and the MOSFET gate, to charge up a cap from the gate to ground, rather than the delay relay in Satoru-san's design. 2.2Meg and 2uF should do it.
It's something to consider, but the use of SS devices in applications like this always scares me. It's possible to achieve very low levels of ripple with even a relatively small choke (1-2 Hy) in a CLC filter. Chokes in that range typically have low DCR and will drop very little DC voltage. Compared to 10-20 Hy chokes, they're also small, light and inexpensive. They don't provide regulation, of course, but I seldom find that necessary for HV. Funny, I was just thinking about this very thing (SS vs choke) as it might apply to voltage doublers. I have a number of old HK doubler transformers I was considering throwing out. A simple means of cleaning up the supplies (CLC or C-SS-C) might make them more useful.
TK,I don't know what kind of H/K power "iron" is on your "pile". However, I DO know what works in a doubler PSU. Follow the general setup in a Jim McShane overhauled H/K Cit 2. Low noise SS diodes into a pair of HIGH value, low ESR, 'lytics. A low DCR choke follows the doubler stack. Follow the choke with additional capacitance of substantial value. The choke primarily kills "hash", not the ripple fundamental. "Hash" comes from SS diode switching noise and ripple overtones related to the highly triangular wave form large caps. yield. DC B+ draw must be limited 0.25X the AC RMS current capability of the rectifier winding.
FWIW, we use the same 820 muF. 'lytics in the "El Cheapo" doubler as Jim uses in Cit. 2s. It's a cost containment measure related to economies of scale. The "El Cheapo" doubler is quite stiff and the ripple freq. fundamental level is low, but there's plenty of overtone "hash". The replacement choke targeted at Dyna ST70s makes short shrift of that crud. Even a "Bargain Basement" project has to have a QUALITY PSU.
Eli D.
My last amp is full wave of 866As into 5+5H, 50uF, 10H, 150uF. I think I get < 5mV on 900VDC. Do I need a mosfet? Yes, that would save money.I used a shunt regulated PSU on my preamp (CCS + gas regulators) ... I switched back to a full passive PSU in a couple of months. I prefer the old way.
Ciao
Gianluca
Why did you move away from the CCS+gas shunt regulator? I was looking at that just yesterday on the driver stage of Lynn Olson's designs and wondering if an IXYS high voltage CCS module would work there.
Thanks
I ended up with a passive PSU to stay simple and because I liked the sound more. Just note I am using a CCS (that IXYS bug) on the anodes of the tubes ... weird? Yes, maybe.Ciao
| ||||||||||||
| ||||||||||||
This post is made possible by the generous support of people like you and our sponsors:
Top of Page ]
[ Contact Us ]
[ Support/Wish List ]
[ Copyright © 2025, Audio Asylum, all rights reserved ]
